Can Hiring Quotas Work? The Effect of the Nitaqat Program on the Saudi Private Sector

Transcription

1 Can Hiring Quotas Work? The Effect of the Nitaqat Program on the Saudi Private Sector Jennifer R. Peck November 2014 Abstract Since 2011, Saudi Arabia has dramatically extended its labor market policies to address youth unemployment and low Saudi participation in the private sector. This paper studies the Nitaqat program, which imposed quotas for Saudi hiring at private firms. The policy provides a unique setting to investigate the effects of quota-based labor regulations on firms. The analysis uses a comprehensive firm-level administrative dataset and exploits kinks in hiring incentives generated by the quotas to estimate policy effects. The findings indicate that the program increased native employment at substantial cost to firms, increasing exit and decreasing total employment at surviving firms. I gratefully acknowledge the support of the Labor Market Decision Support System project at the Center for Complex Engineering Systems at the King Abdulaziz City for Science and Technology in Saudi Arabia and the Massachusetts Institute of Technology. I am extremely grateful to Michael Greenstone and David Autor for their extensive feedback and to Ahmed Fadol for his considerable assistance with this project. I also thank the rest of the LMDSS team for their support and advice. Miikka Rokkanen, Heidi Williams, and seminar participants at MIT contributed many helpful comments and suggestions. Swarthmore College Department of Economics, 500 College Avenue, Swarthmore, PA ( jpeck1@swarthmore.edu)

2 1 Introduction Many countries have used quotas and affirmative action policies to favor members of disadvantaged or underrepresented groups. These policies aim to increase the representation of these groups in a variety of areas, including elected positions, education, and employment [Fryer & Loury 2013, Sowell 2005]. Legislative gender quotas, for example, are quite common worldwide, and many universities either explicitly or implicitly consider race and gender as a factor in admissions. Government-mandated quotas and group preferences are also frequently applied in labor markets, both to civil service positions and to employment at private sector firms. 1 One of the key issues regarding these types of quota-based labor policies is the tradeoffs they impose between their benefits to targeted groups, the costs to other workers, and the impacts on firms. Theoretical models yield ambiguous predictions regarding the efficiency impacts of these policies, and the net effects depend on both the type of discrimination being modeled and the particular labor market context [Holzer & Neumark 2000]. Empirical evidence on the effects of these policies in various settings is therefore essential. This paper examines one of the world s largest quota-based labor policies to estimate the effect of hiring quotas on employment and firms. The Nitaqat policy was enacted in Saudi Arabia in 2011, and requires Saudi private-sector firms to meet specific employment quotas for Saudi nationals. This policy is attractive to study for several reasons. First, the policy was applied to all private sector firms with more than ten employees, making it one of the most broadly applied such quota programs. The quotas were also rigorously enforced, with sanctions triggered automatically for non-compliant firms. Quota compliance was also carefully monitored through the government s integrated social security and visa records. The policy was therefore both clearly-defined and well-enforced. In addition to providing important evidence on the effects of quota-based labor policies, Nitaqat also offers a window into government efforts to combat the effects of the resource curse on the labor market. Many resource-rich countries face a number of common economic problems, including the underdevelopment of the non-resource sector, high unemployment, weak institutions and corruption, and political instability. These challenges were particularly salient for Middle 1 Affirmative action in the United States, for example, applies to government jobs as well as to private firms with government contracts. New Economic Policy regulations in Malaysia and post-apartheid employment equity policies in South Africa apply to both public and private-sector jobs. 1

3 Eastern oil producers during the Arab Spring uprisings of 2011 and 2012, and protests occurred in almost all of the oil-exporting countries in the Middle East. 2 While there were certainly a variety of reasons for the demonstrations, protestors often cited unemployment as a central concern, especially in places like the usually peaceful Oman. Indeed, high unemployment rates, particularly among young people, tend to be a crucial issue for governments worried about political instability. 3 In addition to experiencing many of these general issues, Saudi Arabia is also part of a subset of oil exporters that share several specific labor market features. All of the countries of the Gulf Cooperation Council (GCC) Saudi Arabia, Bahrain, Oman, Kuwait, Qatar and the UAE have economies that are characterized by several common core issues. In particular, all six countries have a very heavy reliance on oil and gas, with fuel exports ranging from 30 to 85 percent of total GDP. 4 The GCC countries also tend to have dramatically segmented labor markets, with large populations of low-skilled migrant workers. These guest workers form between 20 and 80 percent of the total workforce in these countries, and non-citizens make up nearly a third of the total GCC population. Correspondingly, there is also a low participation of nationals in the private sector, with most citizens working in the public sector or in the oil and gas industry. 5 At the same time, these economies tend to suffer from high and rising unemployment, especially among young people. Saudi Arabia is a clear example of this pattern, with a large number of guest workers, high native unemployment, and sluggish growth in the non-oil private sector. Saudi nationals form about half of the labor force, with four million Saudis employed in Of these, sixty percent worked in the public sector, and only about 600,000 worked in the non-oil private sector. Foreign, or expatriate guest workers make up ninety percent of the non-oil private sector workforce. 67 Unemployment is also very high among new labor market entrants, and official figures from the Central Department of Statistics and Information (CDSI) report 40 percent unemployment in the age group. 2 Uprisings or protests were documented in Libya, Yemen, Tunisia, Egypt, Bahrain, Algeria, Syria, Iraq, Kuwait, Morocco, Oman and Saudi Arabia. 3 Because the governments in the Middle East also control between a half and one third of world oil reserves, their political stability is often of great international interest as well. 4 In comparison, Venezuela derives only 18 percent of GDP from oil and gas. 5 Until recently, the GCC states used public sector employment as a way to combat unemployment and redistribute oil wealth. This strategy has become unsustainable as population growth has rapidly outpaced growth in oil revenues [Forstenlechner & Rutledge 2010, Forstenlechner, Madi, Selim & Rutledge 2012, El-Katiri, Fattouh & Segal 2011]. 6 I follow the convention of referring to these migrant workers as expatriates. This terminology is used to indicate both the broad skill spectrum of these guest workers as well as their temporary residence in the country. 7 These expatriate workers form one of the world s largest migrant populations: in 2010 Saudi Arabia was the fourth largest destination for migrants after the United States, Russia, and Germany, with, with 7.3 million immigrants forming a striking 28 percent of its population (World Bank 2011). 2

4 The reliance on migrant labor in the face of rising national unemployment has become a critical issue for Saudi Arabia and the rest of the GCC. Many firms prefer to hire low-cost foreign labor rather than Saudi workers, who are usually more expensive and less flexible. 8 In comparison to nationals, who often have access to generous government benefits and services, expatriates tend to accept lower wages and to work longer hours in poorer conditions. The de facto minimum wage for a Saudi worker, for example, is around 3000 SR (about 800 USD) per month. In contrast, expatriate workers can be paid about 1500 SR per month, or 400 USD. Although expatriates must be recruited from overseas, their employment terms are also much more flexible than those of Saudi employees, who are more difficult to fire under Saudi labor laws. Foreign workers usually come to the GCC without their families and are not offered a path to citizenship; their ties to their host countries remain very loose, and many send their wages back to their home countries as remittances. Throughout the GCC, governments have become increasingly concerned about both rising citizen unemployment and continued dependence on foreign labor. In addition to political concerns about the potential for radicalization among unemployed youth, large expatriate populations themselves are seen by elites as potentially politically destabilizing, making nationalization efforts highly politically desirable [Randeree 2012, Al-Dosary 2004, Al-Lamki 1998]. Over the past thirty years, all six countries have instituted some form of private-sector workforce nationalization program to address these two issues. 9 These programs are known as Saudization in Saudi Arabia, Bahrainization in Bahrain, Kuwaitization in Kuwait, Omanization in Oman, Qatarization in Qatar, and Emiratization in the UAE. These initiatives are the core government strategies to both increase national employment and to reduce dependence on a foreign workforce. Until recently, however, these programs have been relatively narrow in scope and largely unenforced [Randeree 2009]. From 1995 to 2010, Saudi Arabia s nationalization efforts were similar to others in the region, with extremely ambitious Saudization targets that were not enforced on a broad scale, but which had achieved some success in the oil and gas industry and in financial services. 10 In 2011, the Saudi Ministry of Labor began enforcing an updated version of the old nationalization program that had 8 While high-skill workers are also brought in from the West for their technical expertise, the majority of expatriate workers are hired for low-skill work. 9 These programs are summarized in Randeree [2012]. 10 Under the old Saudization law, companies in nine sectors were required to achieve 30 percent nationalization targets, and construction companies were assigned a 10 percent Saudization target. This law was not enforced, however, and companies in most sectors fell well short of these quotas. 3

5 previously been on the books but non-binding. This new program, called Nitaqat, or bands, was designed to give firms more attainable targets and to introduce incentives to achieve nationalization quotas. The program developed nationalization targets based on firm size and industry and imposed visa restrictions based on how firms performed relative to these targets. These incentives have been strictly enforced, and non-compliers have faced restrictions on their work visas for foreign workers, while firms that perform well are given expedited access to Ministry services such as recruiting assistance and visa approvals. This employment quota program is unprecedented in the breadth of its scope as well as its rigorous enforcement and close monitoring. Because of this, the Nitaqat program is a key test case to measure the potential of these programs to combat unemployment. As with quota-based policies around the world, however, one of the main concerns about these programs is that they will overburden an already-fragile private sector [Looney 2004, Ramady 2013, Hertog 2014]. These regulations are expected to raise costs for private sector firms, which could significantly handicap growth even as these countries rely more heavily on this sector to diversify their economies away from oil and gas. 11 The Nitaqat program is also an important case study for how the costs imposed by such a quota program can restrict the growth of the targeted firms. This paper focuses on two main questions: was Nitaqat successful in increasing the number of Saudis in the private sector, and what were the costs to firms? To answer these questions, the analysis employs a comprehensive dataset on the full universe of Saudi private-sector firms used by the Ministry of Labor to administer the program. The data is particularly notable for its wide coverage and high quality, as employment submissions from firms were automatically checked against government social security and visa records. The establishment-level data contains weekly totals of Saudi and non-saudi employees as well as basic firm characteristics such as industry and size category as well as the level of quota compliance. This is the first time that such establishmentlevel data has been made available to researchers, and this access provides a unique opportunity to study the firm-level effects of this program. The main empirical strategy exploits a kink in the incentive to increase Saudization generated by industry-level Nitaqat quotas. This kink in the policy rule yields an identification strategy based on discontinuities in the derivatives of the outcome variables. I use this regression kink design (RKD) to estimate the effect of the Nitaqat 11 The likely economic costs for private sector firms are carefully discussed in Ramady [2013]. The paper also discusses potential interactions of Nitaqat with other proposed government labor market policies. 4

6 program on firms near the quota cutoffs in terms of program benefits (Saudization, Saudi hiring, and expatriate downsizing) as well as program costs in terms of firm size and exit. I also use a differences-in-differences approach to provide an approximate estimate of the overall effects. The analysis finds that the program succeeded in increasing Saudi employment, but did so at significant costs to firm growth and survival. Program compliance rates were significant, with firms increasing their Saudization rate by 0.2 percentage points on average for every percentage point increase required to reach the quota. Quota compliance was primarily accomplished by hiring Saudis, and the paper estimates that Nitaqat was responsible for the addition of 73,000 Saudi workers to existing private sector firms over the 16 month period, a sizable share of the approximately 460,000 new Saudi workers employed at these firms over the period. New entrants also tended to have higher Saudi employment rates, implying an additional 23,000 positions for Saudi workers in these firms. At the same time, the program caused almost 11,000 firms to shut down, raising exit rates from 19 percent to 28 percent. Surviving firms also tended to shrink in terms of the total number of employees, and the program decreased total private sector employment by 418,000 workers. There is also some evidence that a small number of firms were able to game the system by hiring Saudi workers on a temporary basis in order to avoid sanctions. General downsizing does not appear to have been strategic, however, and firms do not appear to have downsized below the size cutoffs for Nitaqat inclusion as a way to escape regulation. The rest of the paper proceeds as follows. Section 2 summarizes some previous work on labor market quota programs. Section 3 describes the structure of the Nitaqat program, and Section 4 describes the data used in this analysis. Section 5 outlines the RKD empirical strategy, its applicability to the analysis of the Nitaqat program, and reports the relevant identification checks. Section 6 reports the main results and some extensions, and section 7 concludes. 2 Background: Previous Literature The analysis of the Nitaqat program relates to a large literature in labor economics on the effects of employment quota programs. The most well-studied of these are affirmative action policies in the United States. 12 Although most of this literature has focused on the effects of affirmative 12 For a detailed survey of this literature, see the comprehensive literature review in Holzer & Neumark [2000]. 5

7 action on employees, there are several studies that have attempted to estimate the effects on firms. Griffin [1992], for example, estimates establishment-level translog cost functions for firms that were government contractors (and therefore subject to affirmative action regulations) and for firms that were not in the contracting sector. He finds that the labor costs of contracting firms were 6.5 percent higher than those of non-contracting firms. In the absence of exogenous variation in which firms were exposed to the regulation, however, it is difficult to know how much of these differences are attributable to affirmative action alone. There are also several recent papers on employment quota programs outside of the United States. Recent studies by Howard & Prakash [2012], Chin & Prakash [2011], and Prakash [2009], for example, have examined the effect of Indian minority hiring quotas on employment outcomes and occupational choice of favored groups. These studies find that these programs increased the probability of finding a salaried job for some types of favored groups, and that this improved employment outcome was associated with higher household consumption expenditures and higher-skilled occupational choice. This study adds to this literature in several ways. First, the strict enforcement and clean colorband assignment cutoffs provide quasi-experimental variation in incentives that allow this study to estimate the causal effect of the quota on firms. This type of evidence is rare in this literature, which often suffers from endogeneity problems in the identification of program effects. This study is the first to examine a quota program of this magnitude, both in terms of the number of industries included in the program as well as its geographical extent. The overall effects of programs that target a particular industry or focus on a single area are likely to be small both because the small number of affected workers as well as the fact that workers may easily be shifted from non-targeted industries or areas. Because of this, the modest effects seen in these types of programs may not be relevant when scaled up to an economy-wide program like Nitaqat. This study will therefore be able to offer a more accurate picture of the effects of a national-level quota policy. This study is also the first to examine the effect of a nationalization policy rather than one targeting a historically disadvantaged minority. The differences in the characteristics of the targeted labor force will also have an effect on the interpretation of these results. This study is also of particular interest given the popularity of nationalization as an employment stimulus program in other resource-rich countries, particularly those in the GCC. All six GCC countries (Saudi Arabia, Bahrain, Oman, Kuwait, Qatar and the UAE) already have some form of 6

8 nationalization program in place [Randeree 2012]. 13 Among these, Nitaqat is unique in its broad scope and its enforcement, and therefore provides an important test case for countries looking to expand their efforts in this area. The UAE, for example, has announced a renewed focus on its Emiratization initiative to bring Emiratis into the private sector. 14 This study adds important evidence to the debate about the efficacy of these programs by providing estimates of both the benefits in terms of the employment of nationals as well as the costs to private sector firms. 3 Background: Saudization and the Nitaqat Program Plans to enact a new nationalization policy were first announced by the Ministry of Labor in early Detailed information about the structure of the program, including specific quotas as well as the corresponding sanctions and benefits for compliance, was released to firms in June Sanctions were phased in starting in September This section discusses some of the potential reasons for low baseline Saudization rates at private sector firms and then presents details on the construction and enforcement of the Nitaqat quotas. 3.1 Baseline Saudization Rates Before the program began, most firms had relatively low baseline Saudization rates, with overall Saudization of 8.7 percent in the 1.1 million firms in the sample in July 2011 (Table IV). This was likely due to a variety of factors, including higher reservation wages for Saudis and lower employment protections for expatriates. In addition, qualified Saudi workers tended to be more difficult to hire than expatriates. This is likely the consequence of limited experience on both the supply and demand side: because of their low engagement with the private sector, Saudi workers may have been less likely to have the required skills, including related work experience and education in relevant fields. Low Saudi employment also meant that many firms had little experience recruiting Saudis and limited access to referral networks. 15 There can also be substantial fixed costs involved in hiring Saudi workers. Saudi women, for example, make up a large fraction of unemployed workers, particularly at higher education levels. However, Saudi law requires that women have 13 Hertog [2014] provides a comprehensive assessment of these and other nationalization efforts in the GCC. 14 R. Kasolowsky, UAE mulls new labor law to attract Emiratis to private sector, Reuters. February 16, There is evidence that these hiring networks are important for recruiting under-represented workers [Miller 2014]. 7

9 physically separate work spaces from their male colleagues as well as separate building entrances. All together, these factors tend to make expatriate labor more attractive for most firms. Although the average Saudization rate was low, there was also quite a bit of heterogeneity in Saudi employment across industries. The share of Saudi workers is highest in industries with jobs that are considered culturally (and legally) acceptable for women and in clerical occupations where the skills are similar to those needed in the public sector. Financial institutions and petroleum and gas extraction, for example, both began with Saudization rates above 75 percent. Industries requiring manual work or specialized skills tended to have the lowest Saudization rates. These included construction, farming, maintenence, transportation, and real estate. In addition to the substantial variation in Saudization across industries, different firms also have very different rates of Saudi employment within industry groups. This is likely due to a mixture of structural and transitory issues. Because employing Saudis requires significant fixed costs, firms that have already made these investments likely find it easier to hire Saudis. These investments may include workspace, developing effective Saudi hiring practices, HR quality, and physical capital to accommodate workers with different types of skills. 3.2 Firm Categories Under Nitaqat, the Saudization quotas that firms face vary by industry and size. All private sector firms are allocated into these industry by size categories based on their economic activity type and number of employees. There are currently 52 different industry categories based on the 3,127 economic activities registered with the Ministry of Commerce. Since June 2011, the program has added 10 new industry classifications, increasing the number of industries from 41 to 52. These new sectors were split off from the existing categories in response to complaints that dissimilar business groups were being held to the same targets. Road cargo transport, for example, was split into long-haul and intra-city trucking. Firms were allowed to change their classification up to one time by appealing to the Ministry of Commerce. Within each category, entities are classified into size groups according to the total number of employees in a single industry category across all branches of the firm. These size groups were set somewhat arbitrarily based on natural economy-wide clustering of firm sizes. 16 The five size 16 Administrators at the Ministry believe that larger firms are seen as more desirable places to work, and therefore 8

10 categories are: very small (< 10 employees), small (10-49 employees), medium ( employees), large ( employees) and giant (3000+ employees). These entity sizes are calculated by the Ministry using data on the number of foreign workers visas held from the Ministry and National Information Center (NIC) records and the number of Saudi employees from the General Organization for Social Insurance (GOSI). The number of Saudi employees is entered as a moving average over thirteen weeks to prevent sharp changes in size category or Saudization percentage. Firms are therefore assigned to industry and size categories according to the economic activity of their branches (as registered with the Ministry of Commerce) and their numbers of employees as calculated by the Ministry of Labor from NIC and GOSI data. For example, a firm with three bakeries with 30 employees at each branch would be counted as a single entity with 90 employees, putting it in the Medium size category. A firm with a jewelry store with 12 employees and a clothing store with 60 employees would be classified as two entities, one Small entity in the Jewelry sector and another Medium entity in the Retail sector. If the firm decided to list as one entity, it would be considered Medium sized with 72 employees, and it would have to achieve Saudization targets for the most stringent sector in which it had any economic activity, in this case the jewelry sector with 20 percent nationalization rather than the retail sector target of 17 percent nationalization. Firms may also be listed as a conglomerate, in which case their business lines are classified as a single entity and coded in the multiple economic activities category. Overall, there are 1.8 million branches and 1.2 million private-sector entities monitored by the Ministry under Nitaqat. 3.3 Nitaqat Color Bands Within each cell of the industry by size classification, firms are assigned to a color group based on their Saudization percentage relative to the Ministry s color group cutoffs for that cell. For a medium-sized construction entity, for example, the color band ranges were: Red: 0-2% Yellow: 2-6% Green: 6-28% Platinum: 28+% A construction firm with 5 Saudi employees and 95 foreign workers would therefore be classified as Yellow with a Saudization rate of 5 percent. Firms with fewer than 10 employees were classified have an easier time recruiting Saudi employees. As such, the targets are almost all weakly increasing in firm size. 9

11 as White and were not included in the program. 17 Industry and size group cutoffs were designed based on pre-nitaqat Saudization rates so that approximately half of firms would be coded as Green or Platinum and the other half as Red or Yellow. The lower bound for the Green band was therefore set approximately at each group s median Saudization level. Cutoffs for the red and platinum bands were set by Ministry staff. A firm s Saudization rate is calculated using as a thirteen-week moving average of the number of Saudi workers registered with GOSI. This smooths shocks and encourages firms to improve their color band status through long-term employment of Saudis rather than through temporary positions Enforcement: Sanctions and Benefits The main services that the Ministry of Labor provides to firms are foreign recruitment and the issuance and renewal of work visas for foreign workers. The introduction of the Nitaqat program coincided with a streamlining of Ministry visa applications in which firms could renew and change their visas online. Firms in the Green and Platinum bands were offered these new expedited visa services, while firms in the Yellow and Red bands faced increasing restrictions over time in their ability to renew existing visas and to recruit foreign workers. In addition to becoming eligible for expedited and more flexible visas for their foreign workers as well as enhanced recruitment services from the ministry, firms in the Green and Platinum bands were also given the ability to offer jobs to foreign workers from the Red or Yellow color band categories. Firms in the Yellow band faced some restrictions on their visa renewals, and were not eligible for the electronic visas or recruitment services. Entities in the Red band could not renew any of their existing visas and were not issued any new visas. Their existing visas were very inflexible, and they were not allowed to open any new facilities or branches. According to the Ministry, the sanctions were designed so that firms that remained in the Red band would find it prohibitively difficult to remain in business. The sanctions and benefits are summarized in Table I along with the timing of their implementation. All sanctions and benefits were being enforced by the end of the first year of the program. 17 There are some exceptions where larger companies are categorized as White in cases where Saudization was not considered feasible. International schools, for example have no Saudi employment quotas. 18 There are several types of non-saudis that can be counted toward the firm s Saudi total, as well as bonuses for hiring disadvantaged groups. The formula is presented in more detail in the data section. 10

12 3.5 Program Results Between July 2011 and October 2012, the number of Saudis employed in the private sector increased by 462,000, and the Ministry has claimed that the program was responsible for the creation of 250,000 jobs for Saudi nationals in its first year. 19 Figure I shows the time series of Saudi and expatriate workers in the private sector. While the number of expatriates in the sector increased by almost the same amount, 467,000, the Saudi workforce grew by 72 percent while the expatriate workforce increased by only 7 percent. There was also a large improvement in firm color-band assignments, with most Red and Yellow firms moving into the Green or Platinum bands by October Table V shows the matrix of firm color band movements, depicted graphically in Figure II. Approximately 50 percent of Red firms improved their status, ending the period in the Yellow, Green or Platinum bands. Approximately 70 percent of Yellow firms improved their status, and very few Green and Platinum firms moved into lower color bands. As expected, the reaction from Green and Platinum firms has been quite positive: an HR representative from a telecommunications company categorized as Green under Nitaqat reports that visa applications are now much quicker and that work visas are easier to obtain. Representatives from companies categorized as yellow and red complained about the prohibitive cost of recruiting and hiring Saudis and the negative effects of visa restrictions driving their business to other GCC countries. A recent article also reports that investors in the Saudi trucking industry complain that Nitaqat has hurt their business, claiming that the restrictions cause them to lose SR 250 million a year for failing to hire enough Saudi truck drivers to meet their 10 percent benchmark Data The primary data for this analysis is the administrative Nitaqat program data collected by the Ministry of Labor. This dataset contains weekly entity-level observations of the employment measures and corresponding color band assignments used by the Ministry to determine program compliance and trigger enforcement measures. The dataset contains firm identifiers including geographic location, industry, size category, and a unique firm identifier. Collected employment

13 measures include counts of Saudi and expatriate employees as well as counts of employees in important groups, such as disabled Saudis, ex-convicts, citizens of other GCC countries, women 21, non-saudi spouses of Saudi citizens, non-saudis with Saudi mothers ( special foreigners), part-time workers, students, and members of displaced tribes from the Rub al Khali with a Saudi passport but no national identity card. For Nitaqat purposes, the Ministry counts non-saudis with Saudi spouses or Saudi mothers, and members of displaced tribes toward the total Saudi employee count. Former Saudi prisoners are equivalent to two Saudi employees, disabled Saudis equivalent to four Saudi employees, and students working part time as half of a full-time Saudi worker. 22 The total number of Saudi workers for Nitaqat Saudization calculations is therefore: Nitaqat Saudis = Saudis + Spouses + Special Foreigners + Gulf Citizens + Displaced Tribes + 2 Ex-Convicts + 4 Disabled Saudis +.5 Students +.5 Part-Time The Nitaqat Saudization rate is calculated as the ratio between this total and the total number of employees. Color band assignments are based on a 13-week moving average of this rate to avoid sudden reclassifications due to temporary staffing shocks. All of these employment measures are updated by the Ministry on a weekly basis using Ministry data on visa issuance for foreign workers and GOSI data on Saudi employment rolls. Data collection began on June 11, 2011, and entities were fully represented and reporting all employees by July 9, Therefore, although data exists for June, all comparisons in this paper are based on a starting date of July 9th. The dataset contains observation up through October 13, The dataset includes observations for over one million firms, 116,873 were large enough to be included in the Nitaqat program at its start in July Of these, 83,568 also appear in the data for October 2012, reflecting exit by 33,305 from the sample ,685 new firms entered over the intervening 16 months. These 83,568 firms form the sample for the empirical analysis of the effects 21 Because employee gender was not used to calculate Nitaqat compliance, this particular series does not appear in the weekly Nitaqat data until January The Nitaqat bonus for former prisoners and disabled workers is applied to up to four employees in each group; subsequent employees in these groups count as one additional employee for Nitaqat purposes. 23 It should be noted that exit from the Nitaqat sample does not necessarily reflect exit from the market, and these exit rates may overestimate overall rates of firm shutdown. Entities may exit the sample by falling below the 10-employee inclusion threshold or by creating a new registration with the Ministry of Labor following a merger, split, or other change in firm structure. 12

14 of the program on employment, size, and firm value among surviving firms. The estimates of the overall effects use the full set of 116,873 baseline firms, with all employment figures set to zero in October 2012 for exiting firms. Analysis of firm exit is also done using the full set of 116,873 baseline firms. These firms are distributed across 37 industries and four size categories at baseline (Table II). These firms appear in 109 of the corresponding Nitaqat industry by size categories. Just over one third of these entities were in the construction industry, with most of these in the smallest size category. Construction firms were also responsible for nearly half of private-sector employment and almost a quarter of Saudi private sector employment (Table III). In addition to being the largest private sector industry, construction also had one of the lowest Saudization rates, with an industry average of 5.8 percent Saudi workers. After construction, the next largest industries were retail and manufacturing, with 20 and 11 percent of the Saudi private sector workforce, respectively. The industry category for conglomerates ( multiple economic activities ) contains a large number of entities, all which have less than 50 employees and which employ less than one percent of the Saudi private sector workforce. Although a large number of firms are exempt from Nitaqat due to the ten-employee inclusion cutoff, the firms included in the program employed over 95 percent of the Saudis and 68 percent of the expatriates in the private sector workforce at baseline. 24 Also of note is the large variation in Saudization rates across industries and within different size groups. In July 2011, Saudis made up less than five percent of the workforce in farming, maintenance, and private labor recruitment services. Financial institutions had the highest starting Saudization rate at 80 percent; petroleum and gas followed at 76 percent, and petrochemicals at 45 percent. Though the total workforce share of firms was roughly declining in firm size, Saudi employment was greater for larger firms (Table IV). Tiny firms accounted for only 3 percent of Saudi employment, small firms for 12 percent, medium firms for 29 percent, and large firms for 37 percent. The 58 giant firms with over 3,000 employees employed only 11 percent of the total workforce and 19 percent of the Saudi workforce. Correspondingly, Saudization rates are higher for larger firms: small firms were only 4 percent Saudi, with less than one Saudi employee per firm in this category on average, and large firms had the highest average Saudization rate of 17 percent. 24 Recent updates to Nitaqat have extended the program to include more of these firms, but the current analysis does not include this time period. 13

15 5 Empirical Strategy The purpose of the empirical analysis is to identify the causal effects of imposing a Saudization percentage quota on firms. In particular, the analysis seeks to estimate how the required increases in Saudization affected (1) actual changes in Saudization (did the program have any effect?), (2) hiring of Saudis and downsizing of expatriates (how did firms achieve their Saudization targets?), and (3) firm size and exit (what costs did these requirements impose on firms?). The policy variable of interest is therefore the compliance requirements that the Nitaqat program imposed on firms, i.e. the amount by which firms were required to increase their Saudization rates to meet their quotas. If these required changes were randomly assigned, the analysis could directly estimate the effect of these requirements on these outcome variables. In this case, however, the policy variable was mechanically determined by the firm s baseline Saudization percentage and the quota for the corresponding industry by size cell. These baseline Saudization rates are potentially endogenous to all of the outcomes of interest; unobserved determinants of baseline Saudization are almost certainly correlated with future changes in the employment of Saudis and expatriates as well as other measures of firm performance. Because of this, the analysis uses the variation in the policy rule generated by the placement of the quotas to identify the causal effects of the program on firms. In particular, the estimation relies on the variation in the incentive to increase Saudization rates created by the quota cutoffs. The Yellow/Green color band cutoffs in particular generated an incentive for firms below the quota (in the Yellow or Red bands) to increase their Saudization rates while imposing no new constraints on Green and Platinum firms with Saudization rates above the cutoff. As discussed below, the quotas generate a kinked assignment function from baseline Saudization percentages to the increase required for program compliance. Because of this, the main analysis uses an RKD to estimate the effects of the program on staffing, firm value, size, and exit. Overall program effects are also estimated using a differences in differences approach comparing the relative changes of Yellow, Red and Green firms within industry and size cells. 5.1 Regression Kink Design The RKD estimates treatment effects using kinks in a continuous policy variable that is based on a potentially endogenous assignment variable. This method is analogous to regression discontinuity 14

16 design, but can be used in cases where the policy variable is continuous but contains discontinuities in its derivative (i.e. kinks). The treatment effect is identified by the discontinuities in the derivatives (i.e. changes in the slopes) of the outcome variables around the kink point in the policy variable. This is critical to the analysis of the Nitaqat program because the hiring incentives approach zero as firms near the cutoff; a Yellow firm with 7.9 percent Saudization facing an 8 percent quota will have almost no need to adjust its staffing. Yellow firms below the quota, however, will need to increase Saudization by an amount that is directly increasing with their distance below the quota, while Green firms incentives to change Saudization rates will be uniformly zero regardless of their distance above the cutoff. The RKD method will allow us to exploit this kink in the quota compliance requirements to estimate the effect of the program near the quota cutoffs. The program s treatment effect will be identified by changes in the slopes of the outcome variables around this kink point in the assignment function. The RKD is formalized by Card, Lee, Pei & Weber [2012], which establishes the conditions under which the RKD identifies the local average response, or treatment on the treated, parameter that would be identified if the treatment had been randomly assigned. The necessary identification tests and robustness checks are similar to those for RDD outlined in detail in Lee & Lemieux [2010]. This method has previously been used for the evaluation of programs with kinked benefit structures such as the EITC [Jones 2013], UI benefits [Card et al. 2012], college financial aid [Nielsen, Sorensen & Taber 2010], intergovernmental grants [Dahlberg, Mörk, Rattsø& Ågren 2008], education finance [Guryan 2001], and prescription drug reimbursement [Simonsen, Skipper & Skipper 2010] Compliance Requirement The RKD analysis in this paper relies on the kinked compliance requirement that was generated by the imposition of Saudization quotas on firms in each industry by size group. As discussed above, the most important quota is the one at the Green/Yellow cutoff. The incentive for these firms to increase their Saudization percentage was increasing in their baseline distance below this cutoff. For example, the cutoff for medium-sized construction firms was six percent; a firm in the Yellow band with four percent Saudi workers needed to increase its Saudization rate by two percent to comply with the program. For Green and Platinum firms above the cutoff, I assume that that firms already in compliance (with baseline Saudization rates just above the quota) experienced no 15

17 incentive to change their Saudization rates as a result of the program. A medium sized construction firm with eight percent Saudi workers, for example, would have a compliance requirement of zero. This generates a kinked function mapping initial Saudi percentage to the increase mandated by the program. Figure IIIa shows this compliance requirement for medium construction firms. This rule generates a similar compliance requirement with a kink at the quota level for each of the 109 industry by size cells; these kinked compliance requirements are plotted for each cell in Figure IIIb. We can combine these by normalizing the cutoff to zero and measuring the compliance requirement as the distance below the cutoff, i.e. b(v ijs ) = max(q js S ijs, 0) where S ijs is the initial Saudization percentage for firm i and Q js is the quota for the corresponding industry j and size group s. For Yellow and Red firms, this will be positive: a firm with a baseline Saudization rate of 5 percent facing a quota of 8 percent would have b(3) = max(3, 0) = 3. A Green firm with 9 percent Saudization facing the same quota would have b( 1) = max( 1, 0) = 0. This normalization collapses the compliance rules in Figure IIIb into a rule with a single kink at zero shown in Figure IIIc. 25 When examining the effect of the program on variables measured in terms of employees, i.e. number of Saudi employees and number of expatriate employees, it will also be useful to define the distance from the cutoff in terms of the number of Saudis that would have to be hired or expatriates that would have to be downsized to meet the quota. For Saudis, we can express this as: Distance S ijs = Saudis ijs Saudis ijs where Saudis ijs : Quota js = Saudis ijs Saudis ijs + Expats ijs. For example, a firm with 18 expatriate employees and 0 Saudi employees facing a quota of Note that this normalization pools firms facing different quota cutoffs into a single sample. This approach is standard in the RD literature when cutoffs vary by treatment site or year (see for example Black, Galdo & Smith [2007]), and yields an estimate of the weighted average effect over cells. 16

18 percent would need to employ 2 Saudi workers to meet the quota, so Distance S ijs = Saudis ijs Saudis ijs = 2 0 = 2 Similarly, for expatriates: Distance E ijs = Expats ijs Expats ijs where Expats ijs : Quota js = Saudis ijs Saudis ijs + Expats. ijs For example, a firm with 12 expatriate employees and 1 Saudi employee facing a quota of 10 percent would need to downsize 3 expatriate workers to meet the quota. These normalizations are useful in the interpretation of the effects of the program in terms of the number of different types of workers employed. The normalized compliance requirements are plotted in Figures IVa and IVb. The assumption that Green and Platinum firms have a compliance requirement of zero is consistent with the idea that baseline Saudization rates reflect unobserved differences in propensity to hire Saudis, whether because of fixed investments made in Saudi HR development, physical capital, or employee-driven recruitment networks. If this propensity to hire Saudis generates an optimal number of Saudi workers that is not affected by the presence of non-binding quotas, then we would not expect these firms to change their staffing in response to Nitaqat regulations. 26 However, this assumption will be violated if firms above the quota experienced pressure to change their Saudi percentages. This may be the case if quotas affected equilibrium wages or resulted in other spillovers from treated (Yellow and Red) to non-treated (Green and Platinum) firms. In this case, firms above the quota would have incentives to move down to the quota, implying a compliance requirement with a smaller kink that the one described above. 27 The results, however, indicate that firms just above the quota tended not to adjust their Saudi employment in response to Nitaqat requirements. 26 If variation baseline Saudization rates are driven by random fluctuations around the median (where the quotas were set), then we would expect Green and Platinum firms to tend to revert to the mean, decreasing their Saudization rates independently of the program. 27 If the compliance requirement was in fact entirely smooth, they the RKD would find no program effect even if the program in fact had a large effect on firms. 17

19 5.1.2 RKD Identification and Estimation The identification assumptions and estimation procedure for RKD are very similar to those required for RDD, but applied to the discontinuity in the derivative rather than the level of the treatment function. In particular, for outcome Y, starting Saudization quota distance V = Q S and Nitaqat compliance requirement B, we can express the effect of the Saudization requirement on the outcome of interest using the generalized nonseparable model Y = y(b, V, U), i.e. define the outcome of interest as a general function of the compliance requirement B, baseline Saudization quota distance (and potentially other observable covariates) V, and an unobserved error term U. The key relationship of interest is the effect of B on Y. 28 The policy parameter of interest is therefore E( Y (B, V, U)/ V V = 0) If B exerts a causal effect on Y and there is a kink in the deterministic relation between B and V at V = 0, then we would expect to see an induced kink in the relationship between Y and V at V = 0. In our case the kink is sharp: the compliance requirement is a deterministic function of baseline Saudization percentage. For a sharp RKD, Card et al. lay out four conditions that must be satisfied for the RKD to identify the causal effects of B on Y. Following their notation, denote the conditional density functions of V on U by F V U=u (v) and f V U=u (v). Let B b(v ) and denote the partial derivatives of y(,, ) by y 1 (b, v, u) = y(b,v,u) b sharp RKD are then: and y 2 (b, v, u) = y(b,v,u) v. The identifying assumptions for the A1: (Regularity) y(,, ) is continuous, and y 1 (b, v, u) is continuous in b for all b, v, and u. The marginal effect of B on Y must therefore be a continuous function of both observables and of the unobserved error term. A2: (Smooth Effect of V ) y 2 (b, v, u) is continuous in v for all b, v, and u. V may affect Y, but the effect is assumed to be continuously differentiable, so any observed kinks in Y cannot be the direct result of small changes in Y. In our case, this would rule 28 In this formulation, the error term U may enter the model non-additively, which allows for unrestricted heterogeneity in the response of Y to V. This setup also allows heterogeneity in the response of Y to B. 18

20 out a kinked underlying relationship between baseline quota distance and increases in Saudi percentage in the absence of the Nitaqat program. A3: (First Stage) b( ) is a known function that is everywhere continuous and continuously differentiable on (, 0) and (0, ), but lim v 0 + b (v) lim v 0 b (v). The compliance function must therefore be known and have a kink at v = 0. There also must be a positive density around the kink point. In our case, the compliance requirement is b(v ) = max(v, 0), so lim v 0 b (v) = 1 0 = lim + v 0 b (v). Because the quotas were placed near the median Saudization rates for each industry by size cell, there is also a large density of firms around this kink point. A4: (Smooth Density) F V U=u (v) is twice continuously differentiable in V for all u and v. This condition rules out the manipulation of the assignment variable and is the key identifying assumption. In summary, if everything else is continuous near the kink, any changes in the slope of the outcome can be attributed to the kink in the compliance requirement B. In this case, the RKD will identify the treatment on the treated parameter at this point, i.e. the average effect of a marginal increase in the compliance requirement near the cutoff holding the distribution of unobservables constant. The degree to which V and U are correlated will determine the extent to which this treatment effect applies to firms that are further away from the quota. There are two testable implications of the identification assumptions above. First, in a valid sharp RKD, f v (v) must be continuously differentiable in v. This rules out precise manipulation of baseline Saudization percentage by firms near the quota cutoffs. This is reasonable given that the quotas were not announced prior to the start of the program: although firms had been informed that the government would start enforcing Saudization quotas, firms were not told where the cutoffs would be for their industry and size groups until the start of the program in June We can test for this by examining the baseline distribution of V. In particular, I use a modified McCrary test to test for a break in the density of V around the kink in the compliance function [McCrary 2008]. Figure V plots the density of baseline Saudization percentages relative to the cutoff. A McCrary 19

21 test shows no evidence of bunching to the right of the quota at the start of the program, and the figure confirms that quotas were set near the median starting Saudization percentages. The second testable implication is that there should be no kink in baseline covariates around the quota, i.e. dp r(x x V =v) dv is continuous in v at v = 0 for all x. 29 Baseline values of several sample covariates (firm size, Saudi employees, and expatriate employees) are plotted in Figure VI; none of these correspond to a statistically significant kink or discontinuity in averages around the cutoff. The fact that quotas were assigned near cell medians also means that there should be roughly the same number of firms above and below the cutoff within industry by size groups. If we use a simple, additive model with constant effects: then, under the above assumptions: τ = lim v 0+ Y = τb + g(v ) + U E[Y V =v] v b(v) lim v 0+ v lim v 0 E[Y V =v] v lim v 0 b(v) v This RKD estimand is the change in the slope of the conditional expectation function E[Y V = v] at the kink point v = 0 divided by the change in the slope of the assignment function b( ) at that same point. In our case, the assignment function is b(v ) = max(v, 0), so the change in the slope of the assignment function is 1 at the cutoff. We therefore have. E[Y V = v] E[Y V = v] τ = lim lim = v 0+ v v 0 v ˆβ 1 where ˆβ 1 is estimated from the model: E[Y V = v] = α 0 + P p=1 [ ] α p v p + β p v p D where v < h for bandwidth h and p is the polynomial order of the fit. The analysis estimates these local polynomial regressions using a symmetric uniform kernel and several values of the bandwidth and the polynomial order. In estimating the derivative at the kink point, Card et al. [2012] point 29 This is analogous to a test for true random assignment in an RCT. 20

22 out that a local quadratic (p = 2) leads to asymptotically smaller bias than a local linear regression (p = 1), but find that this comes at the cost of significantly larger asymptotic variance. 30 While the analysis reports the estimates from both p = 1 and p = 2, the local linear specification is therefore preferred. The results will also be reported for several choices of bandwidth. 5.2 Differences in Differences While the RKD analysis focuses on changes in incentives to hire around the kink in the policy rule, it is also useful to estimate the overall effects of the Nitaqat program on Saudi employment, expatriate employment, firm size and exit. 31 This can be done by estimating the average effect of assignment to the Red or Yellow color bands as compared to firms in the Green band within the same industry by size cell: Y ijs = γ 1 D(Red) ijs + γ 2 D(Y ellow) ijs + α js + ɛ ijs Because Green firms that were well-above the cutoff are likely to have been affected by the program, the analysis also reports estimates where only Green firms within five Saudi employees of the cutoff are used as a comparison group: Y ijs = γ 1 D(Red) ijs + γ 2 D(Y ellow) ijs + γ 3D(Green > 5) + α js + ɛ ijs where D(Green > 5) is a dummy variable for the group of Green firms which began above the quota by more than five Saudi employees. This method requires more assumptions about the effects of Nitaqat on firms above the cutoff than the RKD. In particular, the quality of these estimates will depend on the assumption that firms just above the quota cutoff provide a good counterfactual for firms below and farther above the cutoffs. If baseline Saudization rates are the results of different Saudi hiring propensities related to firm characteriscis (including fixed hiring investments), Green and Platinum firms well-above the 30 They calculate that the bias in the linear specification would have to be 15 times larger than the standard error for a local quadratic to reduce the mean squared error over the local linear regression. 31 In the differences-in-differences analysis, changes in Saudization percentage is calculated only for firms in the matched sample, while changes in employee counts (Saudis, expatriates, and total) are based on all firms in the baseline data. Firms that exit before October 2012 are assigned employment values of zero in the October data. 21

23 quota will tend to be a less useful comparison group than those just above the quota. Spillover effects on Green and Platinum firms will also bias the results. For example, these estimates will be too small if Green firms just above the cutoff took additional steps to retain or hire Saudi workers because of the presence of the quota. 32 The estimates will tend to be too large if Yellow and Red firms met their quotas by poaching employees from Green and Platinum firms. We may also be concerned about other types of market-level spillovers, such as wage effects, competitive effects on exits, or price effects in goods markets. Even with these issues it is nonetheless helpful to get a sense of the magnitude of the overall effects. 33 To account for the effects of Nitaqat on the workforce composition of new entrants, the analysis also estimate how firms that entered after Nitaqat began compare with firms that entered in the first month of the data. In particular, the sample of 40,620 firms that entered between July 31, 2011 and October 13, 2012 is compared with the 5,065 firms that entered in July 2011 in the same industry by size groups. For Saudization percentage, Saudi employees, expatriate employees, and total employees, the specification is: Y ijs = δ D(New Entrant) ijs + α js + ɛ ijs Of course, the quality of these estimates will depend on the relevance of the July 2011 entrants as a comparison group for July 2011-October 2012 entrants in October Results 6.1 Quota Compliance The RKD plots of the change in Saudization percentage against the starting quota distance are presented in Figure VIIa. 35 The program appears to have had at least some of the intended effect 32 The RKD results show that these firms kept their Saudi hiring constant, which is somewhat reassuring. 33 To complement the RKD analysis, I also examine how this effect changes with starting distance from the cutoff. See appendix A for empirical setup and results. 34 Nitaqat may have also affected the number of new entrants in addition to the composition of those entrants. Unfortunately it is not possible to estimate the effect on entry rates due to the limited time horizon of the data. The estimates of the effects on Saudi employment therefore assume that the number of entrants was not affected but that their composition may have changed. 35 Circles plot the average change in the outcome variable for firms according to starting distance from the relevant Nitaqat quota. Lines show the linear regression fit for a bandwidth of

24 on firms below the quota: Yellow and Red firms tended to increase their Saudization percentages to comply with quota requirements. This compliance is particularly sharp very close to the quota, where firms needed to make smaller changes to their employee mix to improve their color band rating. This effect is partially offset by Green firms, however, who tended to reduce their Saudization percentage with increasing distance above their quotas. Although compliance rates are high among Yellow and Red firms, this reduces the estimated effect of the program. The estimates for the size of the kink are reported in Table VII, with most local linear estimates around This indicates that the effect of the program was to increase Saudization in Yellow and Red firms just below the cutoff by 0.2 percentage points for every percentage point they started below the relevant quota. Firms could achieve these increases in Saudization percentage both by downsizing expatriates and by hiring Saudis. Figures VIIb and VIIc show the RKD results for the Saudi and expatriate employment outcomes. There is a clear kink in the number of Saudi hires as a function of the firm s initial distance from the quota in terms of Saudi employees. Yellow and Red firms close to the cutoff hired almost exactly as many Saudis as they needed to reach their Saudization quotas without changing their expatriate worker totals. In contrast, Green firms just to the right of the cutoff experienced no change in their number of Saudi employees. The econometric results in Table VII confirm this, with estimates ranging from 0.35 to 0.59 Saudi workers hired for each one needed to meet the quota. Expatriate workers, on the other hand, show little responsiveness to quota cutoffs, though expatriate hiring increases in distance above the quota. This suggests that firms were not changing their expatriate staffing in order to achieve the quotas. The visa restrictions placed on Yellow and Red firms (and the streamlined renewals offered to Green firms) likely reduced expatriate hiring at Yellow and Red firms while encouraging an increase in hiring at Green firms. Yellow and Red firms far below the cutoff were the least likely to improve their color band assignment and become eligible for the enhanced recruitment services. Similarly, Green firms well-above the cutoff were both unconstrained by quotas and likely to maintain access to visa services over the period. The estimates in Table VII confirm that there is little evidence of a kink, with all of the point estimates close to zero and most not statistically significant. There is also some evidence that Green firms farther above the cutoff tended to reduce their numbers of Saudi employees. 36 This may be because the program made experienced Saudi employees more valuable 36 These findings are supported by the results in appendix A. 23

25 to Yellow and Red firms, so firms well-above the cutoff allowed their employees to be poached by other firms. This effect will mitigate the overall impact of the program on Saudi employment by simply shuffling already-employed Saudis between firms. It is therefore important to allow this group of Green firms well-above the cutoff to be treated by the program in the analysis below estimating the overall program effects. When exiting firms are included (Figure VIII), Saudi hiring rates for Yellow and Red firms farther below the quota are lower, reflecting the higher exit rates among these firms. The results on expatriate hiring are roughly the same Program Costs The effects of the program on firm size and exit are shown in Figures VIId and VIIe. Panel (d) plots the percentage point change in firm size relative to the initial percentage point distance from the Nitaqat cutoff. Firms that remained in the sample over the whole period grew on average, and the intercept indicates a growth of about 25 percent among firms just at the cutoff. On average, firms above the cutoff appear to have grown at about this rate. For Yellow and Red firms below the cutoff, however, the effect on firm size is dramatic, with the growth of these firms dropping off sharply in cutoff distance. The estimates for firm size in Table VII mostly indicate a 0.5 percentage point decrease in firm growth for every percentage point below the cutoff. This indicates that the increase in Saudi employees was not the only effect of the program, and that Nitaqat imposed serious constraints on firm growth over the 16-month period. Although these firms tended to increase their Saudi workforce in response to the program, this result indicates that these firms tended to lose workers overall as their visas were restricted. While firms do not appear to have reduced their expatriate workforce in order to comply with Nitaqat as a function of their quota distance in terms of expatriates (panel (b)), firms that started further from the quota in percentage terms experienced a larger overall reduction in firm size (panel (c)). Once again, the qualitative results are very similar when exiting firms are included (Figure VIII, panel (c)): including these firms in the firm size results shifts the fitted lines downward, but leaves the kink unchanged. Also on the cost side, the Nitaqat program also appears to have significantly increased firm exit. Figure VIIe shows the graphical results, plotting average exit rate against percentage point distance 37 See appendix B for sector and industry-level RKD results. 24

26 from the cutoff. Firms above the cutoff experienced little effect on exit rate, with the average exit rate for Green firms at around 15 percent regardless of cutoff distance. 38 For Yellow and Red firms, exit rates are increasing in distance below the cutoff; although estimates for exit rate are sensitive to the choice of bandwidth, and all linear kink estimates fall in the range. This indicates that exit rates increased by between 0.72 and 2.17 percentage points for every baseline percentage point below the cutoff Overall Effects Estimates for the overall effects of the program are displayed in Table VIII. These estimates are based on cell-level difference-in-difference estimates calculating the average effects by initial color band assignment. 40 Odd-numbered columns show comparisons against all firms in the Green band; even-numbered columns allow for poaching, or changes in Green firms that were more than five employees above the cutoff, by using only Green firms near the cutoff as the comparison group. This effect appears to be important, and the conclusions focus on these results. The last two rows of the table show the total estimated effect of the program based on these estimates as well as the relevant full-compliance benchmark. 41 The table shows that Yellow and Red firms increased their Saudization percentages by 3.5 percentage points on average, with Green firms reducing their Saudization rates by 5.45 percentage points. Overall, the program is estimated to have increased Saudization by 2.73 percentage points, compared to an estimated full-compliance benchmark of On the Saudi employment side, Red and Yellow firms increased their Saudi employment by around one employee on average, while Green firms reduced their employment of Saudis by around nine employees. Because there are many more Red and Yellow firms, however, the overall effect is to increase Saudi employment by 73,000 over half of the benchmark of 138,000 but well short of the no-poaching benchmark of 38 About percent of businesses in the U.S. and the U.K. close each year. 39 A similar analysis was done for the effect on Nitaqat on the market value of publicly-listed firms. Unfortunately the small sample size makes it impossible either to detect a kink in market value changes or to find a sufficiently precise zero. This is also the case for other balance sheet measures available in the Tadawul stock market data for these firms. More information on the data and this analysis is presented in appendix C. 40 As discussed above, these estimates are valid only under strong assumptions about the validity of the firms in the Green band as a control group. 41 In odd columns this benchmark is the change in the outcome variable associated with all firms moving up to the relevant Nitaqat quota, with no change in Green and Platinum firms. In even columns, the benchmark includes the effect of all firms above the quota adjusting down to the quota as well. 25

27 307,000. This implies that Nitaqat was responsible for 16 percent of the total increase in Saudi employment over the period. There is also evidence that Nitaqat reduced the overall size of the expatriate workforce, with Red firms hiring 6.72 fewer expatriates than firms in the comparison group. Expatriate hiring in Green firms increased for firms farther above the cutoff, however, and these increased expatriate employment by on average. Overall, the estimated effect was a reduction in expatriate employment of 496,000, a decrease of just over 50 percent relative to the implied counterfactual increase in expatriate employment in the Kingdom. These effects on Saudi and expatriate employment are reflected in the estimates for the changes in total firm size, implying a reduction in total private sector employment of 418,000 workers. The effects on exit rates were largest for Red firms, with these firms percent more likely to exit than the comparison group. Yellow firms had an average exit rate 4.31 percent higher as a result of the program. Unlike with the other outcomes, Green firms with more than five excess Saudi employees did not experience a differential effect on their exit rates. Overall, the effect of the program was to increase exit by 10,665 firms. This is a significant proportion of the 33,305 firms that exited during the period, implying that the program increased exit rates from 19 to 28 percent. In addition to the effects on firms that already existed when the policy was enacted, Nitaqat quotas also affected the composition of new firms that entered the private sector after July Figure IX plots the distribution of firms in terms of their distance from the quota (in terms of percent Saudization) for new entrants in July 2011 and in October A Kolmogorov-Smirnov test rejects the equality of these distributions, and the distribution of October entrants is shifted to the right: firms that entered after the policy took effect tended to have higher Saudization rates than firms that entered before the quotas were enforced. The average effects for all firms that entered between July 2011 and October 2012 are shown in Table IX. Compared with the firms that entered in the first month of data (July 2011), firm that entered afterward had a 3.76 percentage point higher Saudization rate, and employed 0.57 more Saudi employees. The firms were also larger, employing an additional 1.02 additional expatriates for a total increase of 1.59 employees. Overall, the total effect of the policy on these new firms was to increase the number of Saudis employed by 23,000 and the total number of expatriates by 41, This may be due to the patterns seen in 42 Unfortunately, the short of window of data before the start of the program means that the entrants in July may not be a good comparison group for entrants over the rest of the year, so these estimates are rougher than those in Table VIII and should be interpreted cautiously. As discussed above, this analysis is also not able to account for 26

28 the quota effects on existing firms, which tended to meet Nitaqat requirements by hiring Saudis rather than by replacing expatriates. 43 The combined results indicate a total effect of increasing Saudi employment by 96,000 workers. 6.4 Temporary Saudi Hiring In addition to the costs that were an expected consequence of the program, (increased turnover and downsizing), there are also potential unintended consequences from incentives created by the particular structure of the program. The extent to which firms are able to game the system may account for the less than full-compliance rates as well as suggest some ways in which the program effects are mis-measured. Because most of the increase in compliance is achieved through increasing Saudi employment, one potential concern is that this hiring does not reflect a real, long-term increase in Saudization. One way that firms can avoid hiring restrictions on expatriate workers is to temporarily improve their color band assignment by hiring a large number of Saudi workers when they need to hire more expatriate workers or renew existing work visas. The program rules try to prevent this by assigning color bands based on the 12-week moving average of Saudi employees. Nonetheless, there are occasionally reports of firms hiring large numbers of low-wage Saudi workers for short periods. To get a sense of the magnitude of this effect, I identify firms whose Saudi employment patterns follow a strong cyclical pattern, with temporary hiring booms followed by an increase in expatriate hiring and a sharp decrease in Saudi employment. I do this by flagging firms that were non-compliant (i.e. in the Yellow or Red bands) for most of the period, but that had at one five-week or longer stretch of being included in the Platinum band. This would give the firm access to the expedited recruiting and visa renewal for sufficient time to make use of these services. I restricted attention to firms that had at least one week-to-week increase and decrease in Saudi employment of more than 30 percent. Flagged firms also increased their expatriate workforce on average over the period and did not significantly increase their Saudi workforce, (i.e. all firms had an average increase in Saudi employment of less than 0.5 workers). Of the 113 firms that met changes in the number of new entrants, which may have also been affected by the program. 43 Although existing firms appear to have met quotas by hiring additional Saudis, these firms also tended to get smaller overall due to Nitaqat penalties restricting their visa renewals. This loss of expatriate workers appears to be the result of quota enforcement rather than Nitaqat incentives and is not related to baseline quota distance. New firms, which would not experience these penalties prior to formation, seem to have experienced only the incentive effect of adding Saudis, causing them to be larger. 27

29 these criteria, 18 showed evidence of these strategic temporary hiring booms (Figure X), and 11 of these firms were in the construction industry. Altogether these firms created approximately 250 temporary jobs. This type of blatant manipulation does not appear to be common, although it is certainly possible that more subtle gaming of the system is more widespread. It is likely more common for firms to hire a small number of temporary Saudi workers to switch into the Green band for a longer period of time, for example. Because these changes are relatively small, they likely have little effect on the estimates based on the July 2011 to October 2012 comparisons. 6.5 Downsizing to Avoid Quotas Another way that firms may avoid penalties is by reducing their size below the ten-employee cutoff for inclusion in the Nitaqat program. Because the Ministry re-codes firms when they leave the program, this will be indistinguishable from exit in the data. If firms in the Yellow and Red bands are more likely to downsize in this way, the above analysis will over-estimate the effect of the program on exit (and under-estimate the effect on firm size). If this is the case, we would expect Yellow and Red firms with just over ten employees to exit at a much higher rate than Green firms with the same number of employees. To get a sense of the magnitude of this potential bias, Figure XI panel (a) compares the distribution of firm sizes of all Nitaqat firms in July 2011 and in October There appears to be little change in the distribution of firm sizes, and there is no apparent decrease in the number of firms near the ten-employee cutoff for inclusion in the program. Panel (b) compares the changes in these distributions by starting color, restricting the sample to firms that appeared in the sample at baseline. For both initially compliant (Green or Platinum) and initially non-compliant (Red or Yellow) firms there is a large decrease in the number of firms with between ten and twenty employees. Because this figure categorizes firms based on their July 2011 color band assignment, new entrants are not included; these dips, therefore, may simply reflect a higher failure rate for smaller firms. 44 Indeed, the pattern is similar for both compliant and non-compliant firms, suggesting that the exit of small, non-compliant firms from the data is not driven by downsizing below the size cutoff. This interpretation is supported by panel (c); exit rates follow the same pattern for firms above and below the Yellow/Green cutoff, and there appears to 44 The different distribution in the October 2012 data from panel (a) to panel (b) reflects the fact that smaller firms tended to have a higher turnover rate. This is consistent with previous observations of larger turnover among small firms, e.g. Dunne, Roberts & Samuelson [1989]. 28

30 be no disproportionate increase in exit rate by Yellow firms. Interestingly, the increase in exit rates appears to be relatively homogeneous across firm sizes Conclusion As growing unemployment has led to political pressure, national employment quota policies have become an increasingly attractive potential solution. While these programs promise a quick and visible remedy to citizen unemployment, however, these new labor market regulations are potentially quite costly for firms. The short-term benefits of increasing employment may come at significant cost to long-term economic growth. Recently, political events in many countries in the Middle East have tipped the political economy toward prioritizing short-term stability, and it is likely that these types of quota policies will become more widely enforced in the region. However, there is almost no empirical evidence to suggest what the magnitudes of the costs and benefits of such a program might be, even in the short term. There is a large literature on the effects of affirmative action policies in the United States, but these results have limited applicability to a broad nationalization policy. In particular, affirmative action policies have been applied on a relatively small scale, and have targeted traditionally disadvantaged groups. Nationalization policies differ from these policies on both counts, and both features are likely to have significant implications for the program effects. Further, very little of the work on affirmative action has studied its effects on firms, and identification has been particularly challenging in this subset of the literature. For a large-scale nationalization policy, the effect on firms is critical, with serious consequences for the growth of the often fragile private sector. This paper examines the short-term effects of a nationalization quota policy in Saudi Arabia using quasi-experimental variation generated by the program structure. On the one hand this context is quite specific: Saudi Arabia is unique in many ways, and the Nitaqat program is the first to be implemented on such a wide scale. However, there are many countries with similar labor market features to Saudi Arabia, and there are several features of this policy which make it a good case study. First, the Saudi government devoted significant resources to the program, and it was 45 It is still likely that firms reacted to the ten-employee threshhold even if larger firms did not shrink in order to avoid the policy. Smaller firms, for example, may have avoided growing beyond the ten-employee cutoff. Without data on these small firms it is not possible to estimate the size of this effect. 29

31 implemented quickly and uniformly applied to all private sector firms. Enforcement was strict, and the quality of the administrative data is very high. In contrast to all previously-studied quota policies, both in the United States and elsewhere, it was an economy-wide program, so the results are more relevant to other national-scale programs. The program was also designed with sharp quota cutoffs, which yields identifying variation in nationalization incentives across firms. This paper finds that while the Nitaqat program did increase native employment, it had a significant negative effect on firms. In particular, the policy is estimated to have increased the growth in Saudi employment in existing firms by approximately 13 percent over a 16 month period, adding 73,000 positions for Saudis to the private sector labor force in these firms, and 23,000 positions at new firms. The program also prevented some of the growth in the expatriate workforce, which grew by 496,000 less than it would have in the absence of the quotas. At the same time, the analysis suggests that the costs of constraining the labor market in this way were substantial; the program decreased total employment in the private sector by 418,000 workers and caused nearly 11,000 firms to exit. These costs were not borne equally across sectors, however, and there is some evidence that the increased costs were most damaging to industries in the secondary sector, including construction and manufacturing. This is particularly interesting given that sluggish growth in this sector is one of the typical symptoms of the resource curse. Taken together, the results indicate that the program s quick results in reducing Saudi unemployment have been created at significant costs to firms. However, the program is likely to have important long-term effects as well, which will mitigate some of the short-run costs. In the medium term, firms can adjust their capital investments to decrease the costs associated with employing more high-skilled Saudi labor. More experience and on the job training will also make Saudi workers more valuable to private-sector firms, decreasing the costs associated with employing Saudis instead of expatriates. Over the long-term, the presence of opportunities to work in the private sector will also likely affect the human capital investments of Saudi nationals. Until recently, the primary purpose of post-secondary education was to qualify Saudis for work in the public sector. Increased national participation in the private sector is likely to align education and other human capital investments with the demands of the private sector. The dynamic effects of the program will therefore be at least as important as the short-run impact, and this will be a critical area for future study. 30

32 References Abuzaid, J. [2012], Saudis find salary for truckers low, Arab News. August 7, Al-Dosary, A. S. [2004], HRD or manpower policy? options for government intervention in the local labour market that depends upon a foreign labour force: The Saudi Arabian perspective, Human Resource Development International 7(1), Al-Lamki, S. M. [1998], Barriers to Omanization in the private sector: The perceptions of Omani graduates, International Journal of Human Resource Management 9(2), Arab News [2012], Nitaqat expanded, Arab News. August 13, Black, D. A., Galdo, J. & Smith, J. A. [2007], Evaluating the worker profiling and reemployment services system using a regression discontinuity approach, American Economic Review 97(2), Card, D., Lee, D., Pei, Z. & Weber, A. [2012], Nonlinear policy rules and the identification and estimation of causal effects in a generalized regression kink design, NBER Working Papers 18564, National Bureau of Economic Research, Inc. Chin, A. & Prakash, N. [2011], The redistributive effects of political reservation for minorities: Evidence from India, Journal of Development Economics 96(2), Dahlberg, M., Mörk, E., Rattsø, J. & Ågren, H. [2008], Using a discontinuous grant rule to identify the effect of grants on local taxes and spending, Journal of Public Economics 92(12), Dunne, T., Roberts, M. J. & Samuelson, L. [1989], The growth and failure of U.S. manufacturing plants, The Quarterly Journal of Economics 104(4), El-Katiri, L., Fattouh, B. & Segal, P. [2011], Anatomy of an oil-based welfare state: Rent distribution in kuwait, Research Paper, Kuwait Programme on Development, Governance and Globalisation in the Gulf States. Forstenlechner, I., Madi, M. T., Selim, H. M. & Rutledge, E. [2012], Emiratisation: determining the factors that inuence the recruitment decisions of employers in the UAE, The International Journal of Human Resource Management 23(2), Forstenlechner, I. & Rutledge, E. [2010], Growing levels of national unemployment in the Arab Gulf: Time to update the social contract, Middle East Policy 17(2), Fryer, R. G. & Loury, G. C. [2013], Valuing diversity, Journal of Political Economy 121(4), Griffin, P. [1992], The impact of affirmative action on labor demand: A test of some implications of the LeChatelier principle, Review of Economics and Statistics 72(2), Guryan, J. [2001], Does money matter? Regression-discontinuity estimates from education finance reform in Massachusetts, NBER Working Papers 8269, National Bureau of Economic Research, Inc. Hertog, S. [2014], Arab Gulf states: an assessment of nationalisation policies, GLMM Research Paper 1, Gulf Labour Markets and Migration Programme, Badia Fiesolana, Italy. 31

33 Holzer, H. & Neumark, D. [2000], Assessing affirmative action, Journal of Economic Literature 38(3), Howard, L. L. & Prakash, N. [2012], Do employment quotas explain the occupational choices of disadvantaged minorities in India?, International Review of Applied Economics 26(4), Jones, M. R. [2013], The EITC and labor supply: evidence from a regression kink design, Working paper. Kasolowsky, R. [2013], UAE mulls new labor law to attract Emiratis to private sector, Reuters. February 16, Lee, D. S. & Lemieux, T. [2010], Regression discontinuity designs in economics, Journal of Economic Literature 48(2), Looney, R. [2004], Saudization and sound economic reforms: Are the two compatible?, Strategic Insights 3(2), 1 9. McCrary, J. [2008], Manipulation of the running variable in the regression discontinuity design: A density test, Journal of Econometrics 142(2), Miller, C. [2014], The persistent effect of temporary affirmative action, Working paper. Nielsen, H. S., Sorensen, T. & Taber, C. [2010], Estimating the effect of student aid on college enrollment: Evidence from a government grant policy reform, American Economic Journal: Economic Policy 2(2), Prakash, N. [2009], Improving the labor market outcomes of minorities: The role of employment quota, IZA Discussion Papers 4386, Institute for the Study of Labor (IZA). Ramady, M. [2013], Gulf unemployment and government policies: prospects for the Saudi labour quota or Nitaqat system, International Journal Economics and Business Research X(Y). Randeree, K. [2009], Strategy, policy, and practice in the nationalization of human capital: Project Emiratization, Research and Practice in Human Resource Management 17(1), Randeree, K. [2012], Workforce nationalization in the Gulf Cooperation Council states, Occasional paper no. 9, Center for International and Regional Studies, Georgetown University School of Foreign Service in Qatar. Simonsen, M., Skipper, L. & Skipper, N. [2010], Price sensitivity of demand for prescription drugs: Exploiting a regression kink design, Working Papers 10-1, University of Aarhus, Aarhus School of Business, Department of Economics. Sowell, T. [2005], Affirmative Action Around the World: An Empirical Study, Yale University Press. World Bank Group, Ozden, Ç., Parsons, C., Schiff, M. & Walmsley, T. L. [2011], Global bilateral migration database. 32

34 Figures Figure I: Weekly Totals of Saudi and Expatriate Private-Sector Employees )*+,-."/0">4*?6"7/.8-.9":+6;;6/<9=" (#$" ("!#'"!#&"!#%"!#$" >4*?69" " )*+,-."/0" "7/.8-.9":+6;;6/<9="!" >4<BC/<9" E*;;" E*;;"" >4<BC/<9" >4<BC/<9" :F-?=" :G-;;/H=" Notes: This figure shows the weekly totals of Saudis and expatriate workers in the Nitaqat data. Vertical lines indicate important dates in program enforcement. Figure II: Movements Between Color Bands (July 2011 to October 2012) 100% 90% Percent Switched to Ending Bands 80% 70% 60% 50% 40% 30% 20% Platinum Green Yellow Red Exit 10% 0% Red Yellow Green Platinum Starting Color Band Notes: This figure shows the proportion of firms in each starting category (x-axis) that transitioned into different color bands. For example, most firms in the yellow starting color band moved to the green category, and less than ten percent moved into the red category by October of the following year. 33